Re: nspluginwrapper patch for testing (was: Re: flash10 vs f10)

On Tue, Jun 30, 2009 at 06:38:41PM +0200, Juergen Lock wrote:
> On Sun, Jun 28, 2009 at 11:27:20PM +0200, Juergen Lock wrote:
> > On Mon, Jun 29, 2009 at 12:37:12AM +0400, Boris Samorodov wrote:
> > > Chagin Dmitry <dchagin_at_freebsd.org> writes:
> > > > On Sun, Jun 28, 2009 at 12:39:09PM -0500, Sean C. Farley wrote:
> > > >> On Sun, 28 Jun 2009, Chagin Dmitry wrote:
> > > >> 
> > > >> > On Sun, Jun 28, 2009 at 04:29:49PM +0000, Brandon Gooch wrote:
> > > >> 
> > > >> *snip*
> > > >> 
> > > >> >> I have. I'm using it with about the same level of success as the 
> > > >> >> previous versions of Flash (I need to map a couple of keys to run 
> > > >> >> `pkill npviewer.bin`).
> > > >> 
> > > >> I also have a key mapping to killall npviewer.bin.
> > > >> 
> > > >> > for amd64 try: compat.linux32.maxssiz=4194304
> > > >> >
> > > >> > for i386 try: limit stacksize to 4mb or set unlimited.
> > > >> 
> > > >> Thank you.  I tested viewing "TXT ISLAND" on YouTube.  Works at higher 
> > > >> limits for me such as 32MB but not at my default of 64MB.
> > > >> 
> > > >> Why/how does this help?  At least on i386, it stops Flash from choking 
> > > >> on YouTube videos when switching to HD or jumping within a video.
> > > >
> > > > the reason? foolish waves in a Uli head and stupid bug in flash plugin.
> > > > flash uses pthread_detach() after pthread_join() call.
> > > > glibc allows such behaviour (if stack limit < 40Mb)  piece of shit...
> > > 
> > .oO(Why does that remind me again of
> > 	http://developers.slashdot.org/comments.pl?sid=1063323&cid=26128419
> > ?)
> > 
> > > If that helps then it's a good idea to create a pkg-message for
> > > the port (Juergen Lock is CCed).
> > 
> >  Well i386 doesn't seem to have a sysctl for linux stacksize, can
> > ulimit be used for this too?  Because I can't seem to reproduce this
> > problem on an i386 20090605 vbox head guest with the new ports, suddenly
> > flash behaves even when I click the youtube `watch in hd' button, even
> > multiple times...  (On 7-stable amd64 with f8 the sysctl does help tho.)
> > 
> >  Oh and if ulimit does indeed work for the linuxulator maybe
> > nspluginwrapper should be patched instead?  (Or at least on i386?)
> 
> Ok the following seems to work here on 7-stable amd64 with the default
> compat.linux32.maxssiz setting, please test elsewhere too:
> (maintainer Cc'd.)
> 
New version that redirects possible error messages to /dev/null:
(in case limit is already lower than 32M...)

Index: Makefile
===================================================================
RCS file: /home/pcvs/ports/www/nspluginwrapper/Makefile,v
retrieving revision 1.13
diff -u -p -r1.13 Makefile
--- Makefile	19 Mar 2009 17:28:49 -0000	1.13
+++ Makefile	30 Jun 2009 16:31:29 -0000
_at__at_ -7,7 +7,7 _at__at_
 
 PORTNAME=	nspluginwrapper
 PORTVERSION=	1.2.2
-PORTREVISION=	2
+PORTREVISION=	3
 CATEGORIES=	www linux emulators
 MASTER_SITES=	http://gwenole.beauchesne.info/projects/nspluginwrapper/files/
 DISTFILES=	${DISTNAME}${EXTRACT_SUFX} ${RPMFILE}
_at__at_ -53,6 +53,8 _at__at_ post-extract:
 post-patch:
 	_at_${REINPLACE_CMD} -e 's,/usr/X11R6,${LOCALBASE},g' \
 		${WRKSRC}/src/npw-config.c
+	${RM} ${WRKSRC}/usr/lib/nspluginwrapper/i386/linux/npviewer.orig
+
 post-install:
 	${MKDIR} ${LIBDIR}/i386/linux
 	${INSTALL_SCRIPT} ${WRKSRC}/usr/lib/nspluginwrapper/i386/linux/* \
Index: files/patch-ulimit
_at__at_ -0,0 +1,7 _at__at_
+Index: usr/lib/nspluginwrapper/i386/linux/npviewer
+_at__at_ -1,4 +1,5 _at__at_
+ #!/bin/sh
+ TARGET_OS=linux
+ TARGET_ARCH=i386
++ulimit -s 32768 2>/dev/null
+ . /usr/local/lib/nspluginwrapper/noarch/npviewer